Systematic Reassessment of an Agriochoerid Oreodont from the Hancock Mammal Quarry, Clarno (Eocene, Duchesnean), Oregon

ABSTRACT 
 The 31 crania of the agriochoerid oreodont from the Hancock Mammal Quarry (HMQ) possess character states unique among described agriochoerid species. Characters unique to the HMQ agriochoerid include a triangular P2; a P3 with a posterolingual cone and heart-shaped occlusal outline; a P4 with separate anterolabial and posterolabial cones, but without a posterolingual conule; upper molars with large styles and lacking paraconules; lower molars without lingual cingulids, but possessing small metastylids and parastylids; and a mandibular symphysis that is perpendicular to the tooth row without ventral displacement of the p1. We subjected previous size comparisons to an F-test and Welch's t-test and found that the tooth row lengths of the HMQ agriochoerid have significantly different variance and means from Agriochoeus matthewi crassus (P = .005), a taxon to which this population has previously been assigned. Taken together, these characters are diagnostic of a new species, which we describe here as Diplobunops kardoula. The HMQ specimens show that some of the diagnostic dental characters of agriochoerids currently in use change with wear and may be biased by taphonomy. Furthermore, characters diagnostic of both Agriochoerus and Diplobunops are present in the HMQ agriochoerid, and many of the species-level diagnostic characters of agriochoerids are continuously variable or polymorphic within the HMQ population.


INTRODUCTION
Discovered in 1955 by amateur paleontologist Lon Hancock (Pratt, 1988), the Hancock Mammal Quarry (HMQ) is a middle Eocene locality in the Clarno Formation, located just outside the town of Fossil in north-central Oregon. The locality is unconformably capped by the ignimbrite of member A of the John Day Formation (39.2 § 0.03 Ma; Bestland et al., 1997), which lies about 5 m above the highest excavated specimens (Hanson, 1996;Bestland et al., 1999). Many of the species present are range extensions of existing species or are previously unknown species, leading to disagreements about the precise biostratigraphic placement of the site (Hanson, 1996;Lucas et al., 2004;Lander and Hanson, 2006;Mihlbachler, 2007).
The mudstone, conglomerates, and minor sandstone that make up the HMQ are 10 m thick (Hanson, 1996). The bone-bearing deposits of the main locality are only 5.3 m thick, with conglomerates at the bottom grading upward to silty and sandy deposits at the top (Hanson, 1996). The most productive excavation of the locality is confined to an area 5.5 m by 11.3 m in surface extent, and although bones are found throughout the deposits, most specimens were removed from the bottom 0.6 m (Pratt, 1988;Hanson, 1996). Unlike other well-known Clarno localities, botanical specimens are rare and poorly preserved at the HMQ; of the 204 paleobotanical specimens collected before the quarry closed in 1974, only 39 were identifiable (McKee, 1970;Pratt, 1988). Of the carbonized seeds and nuts identified, most belonged to genera currently living in tropical to subtropical environments (McKee, 1970). Data from correlative paleosols also indicate a humid subtropical climate, and the depositional environment of the locality is interpreted as a point bar along a meandering stream (Pratt, 1988). The recovered fauna consists largely of macrovertebrates (more than 10 kg), including the endemic species Eubrontotherium clarnoensis Mihlbachler, 2007. Few vertebrate microfossils were ever recovered at the HMQ, although the site has never been screenwashed (Pratt, 1988).
Of the approximately 2000 vertebrate specimens recovered, 41 were identified as belonging to an agriochoerid oreodont, initially assigned to Diplobunops Peterson, 1919, because of its exceptionally large canines and expanded premaxillae (Hanson, 1996). Lucas et al. (2004) later suggested that the HMQ specimens were a new species, although the species was not formally described. Lander and Hanson (2006) considered the characters of an expanded rostrum and undifferentiated P4 parametacone insufficient for a separate genus and placed all previously described Diplobunops specimens into Agriochoerus Leidy, 1850. Lander andHanson (2006) described the HMQ specimens as a new subspecies of Agriochoerus matthewi (Peterson, 1919), A. m. crassus. In an unpublished M.S. thesis, Ludtke (2008 suggested the resurrection of the genus Diplobunops on the basis of a rostral constriction at P2, a large C1-P1 diastema, and several tooth characteristics that link the genus with Protoreodon but separate it from Agriochoerus. Ludtke (2008) assigned the HMQ specimens to D. crassus, a species he separated from Agriochoerus by the rostral constriction at P2 and several other tooth characteristics, and characterized it as larger than D. matthewi. Here we reevaluate that assignment based on a reexamination of the sample.

METHODS
All measurements were taken using Mitutoyo Absolute Digimatic digital calipers. All measurements were taken to produce the maximum value. To quantify intraoperator variability, five UOMNH specimens were remeasured three times each, with an hour between measurements. In addition to the use of original drawings and descriptions, we examined photographs of specimens of D. crassus and D. matthewi for potential new diagnostic characters.

Terminology and Abbreviations
Our dental nomenclature follows B€ armann and R€ ossner (2011), save for 'paraconule,' 'posterolingual conule,' and 'preentostylid,' which describe features not discussed in that paper (Fig. 1). The posterolingual conule (hypoconule of Ludtke, 2008 andLander andHanson, 2006) describes an accessory cusp on the posterior side of the lingual cone of the fourth premolar. Preentostylid is used here to describe the stylid on the anterior of the entoconid, which is separated from the metastylid.
'Oreodont' is used here as a general term for any member of superfamily Merycoidodontoidea, whereas 'agriochoerid oreodont' specifies a member of the family Agriochoeridae. Our specimens come from the University of Oregon's Museum of Natural and Cultural History (UOMNH) and the University of California's Museum of Paleontology (UCMP). For UCMP numbers, the locality number is attached and comes first, the catalog number second; for UOMNH numbers, the catalog number comes first and the locality number second.
Revised Diagnosis-The rostrum is constricted at P2, as in all species of Diplobunops but unlike Agriochoerus. The P3 differs from D. crassus and D. matthewi in having a mesolingual cone that is adjusted posteriorly and in possessing strongly pronounced anterior and posterior styles that give it a heart-shaped appearance. The P4 differs from D. matthewi and D. crassus in having fully separated anterolabial and posterolabial cones, but differs from Agriochoerus in lacking a posterolingual conule. The upper molars differ from other species of Diplobunops and Protoreodon in lacking paraconules. The upper molars differ from D. crassus and D. matthewi in possessing very large, rounded meso-and metastyles. Unlike in D. crassus, the symphysis of the mandible does not jut anteriorly, and the p1 crown is not bent to labial. The lower molars differ from those of other species of Diplobunops in lacking lingual cingulids and having small metastylids and parastylids.
No specimens have teeth or premaxillae anterior to C1, precluding description of incisors. The zygomatic arches and postorbital bar were not preserved in any specimens.
The canines are enlarged and 'D'-shaped in occlusal view, with a flat mesial surface. Their height is highly variable, possibly indicative of sexual dimorphism. The rostrum is expanded at the canines and constricted again at P2.
The P1 is bladelike and shaped like an equilateral triangle in mesial view in specimens without wear. It is oval in occlusal view, and the anterior end is bent to labial, putting it out of alignment with the tooth row but maintaining alignment with the constricted muzzle. Although a small diastema between P1 and P2 is typical in our sample, one adult specimen (UOMNH F-27967) lacks this diastema, demonstrating some variability in this character. The P2 possesses a weak mesolingual cone, giving it a triangular shape unlike the rectangular P2 of D. crassus (Scott, 1945;Ludtke, 2008). The P3 has strong anterior and posterior styles and is heart-shaped in occlusal view. There are no crests connecting the mesolabial cone to the mesolingual cone on P3. The P3 mesolingual cone is posterior to the centerline of the tooth.
The P4 has separate antero-and posterolabial cones but lacks a posterolingual conule (Figs. 1A, B and 2A). The anterolabial and posterolabial cones of P4 are close together, so that greater tooth wear gives them the appearance of a partial, rather than full split. This difference can be seen in UOMNH F-27697 ( Fig. 1) or UCMP V75203/218678 (Fig. S1) specimens with demonstrable P4 wear, compared with UOMNH F-30334, which displays two less worn cones (Fig. 2). The P4 is an isosceles triangle in outline when viewed from the occlusal surface, with the mesolingual cone lying almost directly on or slightly anterior to the centerline of the anterolabial and posterolabial cones. The P4 has an accessory crest on the mesolingual cone that becomes more prominent with wear.
As in all oreodonts, the molars are incompletely selenodont, the postprotocrista (posterior arm on the protocone of Ludtke, 2008) and premetaconule crista (the anterior arm of the metaconule of Ludtke, 2008) do not form a loph (Ludtke, 2008;Thorpe, 1937). Paraconules are not present. A cingulum of variable strength is present on the anterior side of the protocone. A cingulum of comparable size is also present on the metaconule of some individuals (Table 1). A cingulum on the anterior face of the protocone of the molars is always present, although not always strongly expressed. Additionally, cingula on the posterior edge of the metaconule and protocone and on the anterior edge of the metaconule are sometimes present. No cingula were found on the labial edge of the upper molars.
The M1 is distinctly smaller than M2 and M3 in both width and length, whereas M2 is only slightly smaller than M3 (Table 2; Fig. 1A, B). The M1 and M2 have large, well-rounded para-and mesostyles, but the metastyle slopes lingually and mesially. The M3 has large, well-rounded para-, meso-, and metastyles, but no prominent labial cingulum overlapping the external fossettes as seen in D. crassus (Scott, 1945). On all three molars, the paraand mesostyles protrude so much that we took no labiolingual measurements on specimens without them to prevent underestimation.
Observations of the cingulum on the anterior side (A) of the protocone, and of the cingulum dam between the protocone and metaconule, or of the posterior (P) side of the protocone in P4 were recorded. 0, not visible; 1, visible but not prominent; 2, prominent, but not shelf-like; 3, shelf-like and very prominent.
Of the mandibles recovered, none preserved canines or incisors. One specimen (UCMP V75203/154600) preserves the entire length of the dentary. The body of the dentary shows little change in depth along its length, and the anterior edge is perpendicular to the tooth row, unlike the anterodorsally projecting symphysis of D. crassus ( Fig. 3; Scott, 1945:pl. V, fig. 1). The ramus is broad and well rounded, much like other Diplobunops species, and possesses a smaller coronoid process and larger posterior condyle than those described of D. crassus (Scott 1945) (Supplementary Data, Table S2).
Only one complete p1 was found, an isolated specimen (UOMNH F-20483) showing the caniniform morphology typical of Merycoidodontoidea. Although the cusps of the p1 of UCMP 154600 were not preserved, enough of the base is retained to show an absence of outward ventral rotation in the orientation of the tooth, contrary to what is seen in D. crassus. No specimens had a p2 with preserved cusps. There is a large p1-p2 diastema in UCMP V75203/154600 but without further samples, it is impossible to determine whether this morphology is typical. The p3 is generally bladelike, non-molariform, and possesses a circular connection of the posterolabial and posterolingual cristids, unlike D. crassus (Ludtke, 2008). There is no distinct mesolingual cone on the p3. The p4 is bladelike when viewed from the side but bowtie-shaped in occlusal view; it lacks an entoconid and is approximately the same size as p3. There is a strong transverse cristid connecting the mesolingual and mesolabial cones of p4 (Fig. 1C, D). The anterolabial cristid wraps around but does not form a complete circle with the anterolingual cristid of p4. The p4 posterolabial and posterolingual cristids both terminate at the posterior edge of the tooth and do not connect.
The metaconid and entoconid of the lower molars are conical, standing perpendicular to the occlusal surface (Fig. 3). There are faint cingulids on the anterior faces of the protoconid and hypoconid of m2 and m3. On m3, the hypoconulid bears a prominent ridge that connects with the entostylid and the posterior arm of the hypoconid. No lower molars show the prominent lingual cingulids or molar ribs that are found in other species of Diplobunops (Lander and Hanson, 2006;Ludtke, 2008). Lower tooth measurements are provided in Supplementary Data, Table S3.

DISCUSSION
One of the key qualitative characters of the HMQ agriochoerid is the split labial cone of the P4, often referred to as the level of 'molarity' (Lander, 1998;Lander and Hanson, 2006). Lander (2006) described the P4 labial cone of the HMQ agriochoerid as minimally split. All of our specimens show two complete P4 labial cones. Theodor (1999) suggested that many of the apparently polymorphic states of Protoreodon species (including the splitting of the P4 labial cone) resulted from grouping species from different stratigraphic levels. Her examination of P. walshi showed no such variability (Theodor, 1999). It is likely that the previously recognized variability of this character in the HMQ agriochoerid came from examining it together with specimens from other localities. Similarly, our examination of the material revealed no paraconules on any of the molars of the HMQ agriochoerid. We suspect that the description of the paraconule in Lander and Hanson (2006) was also a consequence of combining specimens from more than one locality into a single species, misinterpreting what we regard as between-species variability in this character. Lander and Hanson (2006) previously grouped the HMQ agriochoerid with two specimens of A. m. crassus (Carnegie Museum of Natural History specimen 2967 and Texas Memorial Museum specimen 42254-6), providing estimated measurements of the P1-M3 and P2-M3 tooth row of the HMQ specimens as additional support for subspecies status. We subjected the measurements provided in Lander and Hanson (2006) to an Ftest and found the difference in variances between the HMQ agriochoerid and other A. m. crassus specimens to be significant at a D 0.05. Consequently, unable to pool the variances in the two groups, we used Welch's approximate t-test for differences between the distribution of tooth row lengths of HMQ specimens and the other individuals of A. m. crassus (Welch, 1947). We found a significant difference between their means as well, allowing us to reject the null hypothesis of a single source population (P < 0.005 for both P1-M3 and P2-M3 tooth row comparisons). Although not conclusive evidence for two species, this significant difference means that size alone indicates that these two samples are substantially different. This quantitative difference is particularly important when placed in the context of the additional morphological differences between the HMQ samples and other specimens assigned to A. m. crassus discussed in the Systematic Paleontology section of this paper.

Trait Variability
Lander and Hanson (2006) noted medial displacement of the P1 in the HMQ specimens (Fig. 3B). This P1 displacement results from the P1-P2 constriction of the muzzle, which is a diagnostic character of Diplobunops. Consequently, these two characters cannot be considered independent and should not be coded separately in future phylogenetic analyses. Additionally, rotation of teeth, particularly the premolars, is a welldocumented defect among modern artiodactyl species and is typically identical on both sides (Miles et al., 1990). Because of both the character linkage with muzzle constriction and the polymorphy of this character in modern artiodactyls, we reject the use of P1 displacement for diagnosing any agriochoerid taxa.
Cingula and cingulids are frequently referred to in descriptions and occasionally used for formal diagnosis within both families of oreodonts (Lander, 1998;Ludtke, 2006Ludtke, , 2008. The lingual cingulids of the lower molars described and figured in Scott's (1945) description of D. crassus were not present in our specimens, so we examined other cingula on upper premolars and molars for their potential as diagnostic characters. Within the small HMQ sample present at the UOMNH, we found that the cingula of the upper molars were variably expressed, their prominence ranging between large enough to form a small shelf to only faintly visible ( Fig. 4; Table 1). Specimen UOMNH F-27687 possessed all three of these cingula and had the strongest relative expression of all our specimens, implying that there may be a single developmental regulatory mechanism for the cingula in the upper molars. In anthracotheres, the cingula and the styles of the teeth show greater morphological variability than other aspects of the tooth (Ducrocq et al., 1995). The variability in the HMQ agriochoerid sample suggests that relative strength of cingula or cingulid expression should not be used for diagnosis in oreodonts, but presence or absence of certain cingula may be consistent enough for future taxonomic use.

Genus Assignment
The diagnosis of genera within the Agriochoeridae has been heavily debated, and specimens (and species) are regularly reassigned between the three named genera (Agriochoerus, Diplobunops, and Protoreodon) and a variable number of unnamed genera (Lander, 1998;Lander and Hanson, 2006;Ludtke, 2008). Our specimens display characters previously considered diagnostic of two of the three named genera, Diplobunops and Agriochoerus, as summarized in Table 3. Of 12 characters commonly used to diagnose agriochoerid genera, the HMQ specimens share seven with Diplobunops, five with Agriochoerus, and one with both (as the P1-P2 diastema is variably expressed in this sample). The only truly discrete morphologies between these two genera that are clearly presented in our specimens are the expanded rostrum and the split P4 labial cone without a P4 posterolingual conule-both characteristics that place the HMQ  specimens with Diplobunops, even as the continuous characters suggest Agriochoerus.
Diplobunops is known from the late Uintan to the early Duchesnean, whereas Agriochoerus is known from the middle Duchesnean to late early Arikareean (Lucas et al., 2004;Ludtke, 2008). Lander (1998), Lander andHanson (2006), and Ludtke (2008) have proposed that Diplobunops represents an early grade phylogenetically below Agriochoerus, a hypothesis supported by the affinities of the HMQ sample with both genera. Protoreodon also shares many genus-level diagnostic characteristics with both Agriochoerus and Diplobunops, leaving it unclear whether the HMQ specimens represent an evolutionary transition or taxonomic confusion between Agriochoerus and Diplobunops.

Possible Sexual Dimorphism
Specimen UOMNH F-27967 displays several characters unique among the HMQ agriochoerid specimens: it lacks a P1-P2 diastema and has a canine that is wider labiolingually but shorter in height than other canines measured ( Table 2). The distance between the canine and the first premolar (13.77 mm) is within 2 standard deviations of the mean of the HMQ sample (12.28 [mean] § 0.88 [SD] mm), whereas the size of each individual tooth is within 1 standard deviation. Sexual dimorphism of the diastema and canine has been previously hypothesized in other oreodonts (Schultz and Falkenbach, 1968;Stevens and Stevens, 1996). Cranial sutures are not preserved well in any of the HMQ specimens, leaving only tooth eruption and wear for ontogenetic age estimates. The molars in UOMNH F-27967 are fully erupted but show very little wear, indicating that this specimen is a young adult. Without a larger sample size, it is impossible to tell whether the reduced size of these characters in UOMNH F-27967 are expressions of sexual dimorphism or individual variability.

CONCLUSIONS
The distinct characters we recognize as diagnostic of a new HMQ species were either overlooked or considered polymorphic traits by previous workers. When examined as a unique population, the diagnostic characters of the HMQ agriochoerid are monomorphic, numerous in type, and together show no overlap with known species of agriochoerids. Therefore, we recognize the HMQ agriochoerid as a new species of Diplobunops, characterized by fully split labial cones of the P4 without posterolingual conule, triangular P2, heart-shaped P3 with a posterolingualal cone, large, rounded styles on the molars, no paraconules, a mandibular symphysis nearly perpendicular to the tooth row, and no lingual cingulids on the lower molars.
Many of the characters used in the past to diagnose agriochoerid taxa are variable within the HMQ agriochoerid, including molar ribs and the P1-P2 diastema. The variation and continuous nature of these characters makes them difficult to use effectively and consistently between workers. Character variability is an important consideration when assigning specimens to taxa at any level. Oreodont taxonomy has historically been plagued by extensive reassignment of specimens between and among subspecies-, species-, genus-, and family-level designations, with different researchers lumping, splitting, or reassigning depending on which characters they found important (Cope, 1884;Thorpe, 1937;Schultz and Falkenbach, 1968;Stevens and Stevens, 1996;Lander, 1998;Ludtke, 2008). Because the HMQ sample is dominated by cranial material and established diagnostic characters for agriochoerids are predominantly cranial, we have not described associated postcranial material here. However, postcranial material could be an important reservoir for additional phylogenetic characters (Klein et al., 2010;Louys et al., 2012), so their study at the HMQ and other oreodont-rich sites should be a priority for future work.
Oreodonts have no modern descendants but are commonly considered analogous to pigs, sheep, or camels (Peterson, 1919;Thorpe, 1937;Scott, 1945;Schultz and Falkenbach, 1968;Stevens and Stevens, 1996;Lander, 1998;Ludtke, 2008). To truly understand the appropriate range of character variability within an oreodont species, greater consideration needs to be paid to modern analogues and how they inform our character choice. Empirical evaluation of agriochoerid taxonomic characters, combined with a thorough examination of analogous and homologous characters in modern artiodactyls, will clarify the use of these characters in taxon diagnosis.
We found that many of the characters previously used to diagnose genera of agriochoerids were not unique to a particular genus, indicating a need to reevaluate the distribution of characters among agriochoerid genera. The only currently available phylogeny of agriochoerids (Ludtke, 2008) found that Agriochoeridae was not monophyletic when Protoreodon and Diplobunops were included, further supporting a needed revision of diagnostic characters in this clade. Diplobunops kardoula could represent an evolutionarily transitional form between Agriochoerus and Diplobunops, although more work on its evolutionary affinities will be required to test this hypothesis. A reassessment of the phylogenetic relationships of agriochoerid oreodonts is outside of the scope of this paper, but remains a necessity before the evolutionary place of the HMQ specimens will be fully understood.  3. Character matrix for genus-level characters of Diplobunops and Agriochoerus, as described in Ludtke (2008), Lander and Hanson (2006), and Theodor (1999 Museum of Natural History, as well as papers and assistance on many occasions. The disparity between these specimens and known agriochoerids was first noticed by P. A. Holroyd (UCMP), and this project would not have been conducted without her observations. C. Schierup (JODA) and P. A. Holroyd gave access to collections, as well as providing photos. D. Strauss (UCMP volunteer) and N. Famoso (U. Oregon) took excellent photos of the Berkeley specimens for this paper. E. B€ armann and G. R€ ossner's 2011 terminology paper guided and greatly clarified the descriptive aspects of this paper. Anonymous reviewers provided invaluable feedback on terminology in this paper as well. The vertebrate paleobiology group at University of Oregon provided important discussion and commentary in the completion of this research.